Accuracy Improvement of Pose Estimation in A Vision Based Algorithm for Landing an Unmanned Helicopter

In this study, a vision-based algorithm for landing unmanned helicopter is developed to increase the accuracy of pose estimation. In the current algorithm, the reference image of helipad has been matched with the image acquired by the UAV mounted camera using SIFT technique to associate between the corresponding feature points of these two images. Then, the corresponding features are used to estimate the pose of the helicopter relative to the helipad using decomposition of Homography matrix. Applying a sensitivity analysis to pose estimation accuracy, we found the important role of spatial distribution of features. So, a method is suggested for choosing the best features in Homography matrix extraction. Moreover, a proper landmark is designed for helipad such as its features have an inherently appropriate local distribution. This method is assessed implementation in MATLAB software via the real images and different scenarios designed in VRML environment are used. The results of proposed algorithm on produced images from different landmarks, improved the pose estimation error about 30 percent.